fix batching bug

local_evaluation.py

@@ -37,9 +37,8 @@ def get_responses(agent, test_data, BATCH_SIZE):
         for turn_id in range(7):
             batch_inputs = [test_data[i][f"turn_{turn_id}"] for i in batch_idx]
             responses = agent.generate_responses(batch_inputs)
-            for resp in responses:
-                for bi in batch_idx:
-                    all_responses[bi][f"turn_{turn_id}"] = resp
+            for bi, resp in zip(batch_idx, responses):
+                all_responses[bi][f"turn_{turn_id}"] = resp
     return all_responses
 
 def evaluate(responses, test_data):